🛡️ Building an Enterprise AML/CFT Risk Intelligence Platform with Gemma 4

A developer built Risk-Gemma4-Agent, an open-source AML/CFT risk intelligence dashboard that pairs automated transaction analytics with semantic compliance reasoning via the Gemma 4 model (gemma-4-26b-a4b-it) and LangChain vector RAG stores, aligned to the Reserve Bank of India's Master KYC Directions. The Streamlit application uses a modular, decoupled architecture that separates stateless analytical compute (gemma4_engine.py) from role-restricted presentation layers, with sidebar-based session authentication. The developer argues open-weight models let regulated banks keep sensitive customer transaction data off third-party cloud APIs.

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine https://dev.to/challenges/hacktoberfest-week4-2026-10-26 I built Risk-Gemma4-Agent , an enterprise-grade Anti-Money Laundering AML and Counter-Terrorist Financing CFT risk intelligence dashboard panel. In the spirit of the Ghost in the Machine theme, this system acts as a high-vigilance digital watchdog that uncovers "invisible footprints" and hidden financial patterns—such as micro-smurfing networks and automated velocity layering—that would otherwise haunt institutional ledger networks undetected. The application features a modular, multi-tab Streamlit architecture with an active, real-time workspace. It includes: You can view the full decoupled system, codebase architecture maps, and Docker setup instructions here: An enterprise-grade Anti-Money Laundering AML and Counter-Terrorist Financing CFT risk intelligence dashboard panel. This system couples high-performance automated data processing with deep semantic compliance reasoning via the Gemma 4 LLM architecture gemma-4-26b-a4b-it and LangChain Vector RAG stores , adhering strictly to the Reserve Bank of India RBI Master KYC Directions. The platform is engineered using a modular, decoupled architecture separating stateless analytical data compute layers gemma4 engine.py from dynamic, role-restricted presentation layers app.py and tabs/ . The global session management authentication loop is handled persistently through the left sidebar console. graph TD %% Global Session Management Authentication Layer subgraph Sidebar Panel Session Gateway Auth 🔑 sidebar: Enter Gemini API Key -- |Validates API Key| IAM Gate{🔐 sidebar: Session Active?} IAM Gate -- |iam logged in == False| Login 🔓 Choose Identity & Click Login Login -- |Mounts session state active iam user| Workspace Unlock 🟢 … The platform separates stateless backend intelligence from dynamic presentation layers: In highly regulated banking environments, financial entities are strictly prohibited from streaming sensitive customer transaction logs over public, third-party cloud APIs. Open innovation and open-weight architectures change the game entirely: